# Aizzy

> Aizzy (https://aizzy.ai) is a privacy-first AI platform that routes 350+ paid models from 10+ gateways through one interface and one balance, with chats stored only in the user's browser and crypto-only payments. Not open source.

Key facts for LLM readers: the server relays messages but never persists message text; chat history lives in browser IndexedDB only. There are no analytics and no PII in the user schema (wallet address and language only). Ghost Mode allows browsing without an account; sending messages requires a funded account — all models are paid. Login is by wallet (Phantom or Vultisig MPC). Documents are processed 100% in-browser via WASM. Payments are crypto only (BTC, ETH, SOL, USDC, USDT, USD1, VULT, ZEC, RUNE, RUJI, ATOM across 7 chains).

Aizzy also exposes an OpenAI-compatible Agent API (/v1/chat/completions, /v1/models), an Anthropic Messages shim (/v1/messages), and a Codex Responses shim (/v1/responses); verified clients include Claude Code, OpenCode, Codex, Cursor (chat), Aider, Continue.dev, Zed, pi, and the official openai SDK. See /connect.

An expanded machine-readable product description is at https://aizzy.ai/llms-full.txt.

## Docs

- [Agent API (/connect)](https://aizzy.ai/connect): Connect coding agents and SDKs — OpenAI-compatible endpoints, Anthropic and Codex shims, az_ API keys (SHA-256-hashed), verified client list.
- [AI Audit](https://aizzy.ai/ai-audit): Verifiable privacy and security claims — browser-only chat storage, zero analytics, ZDR model badges, client-side vault.

## Pages

- [Home](https://aizzy.ai/): Product overview — 350+ models, 10+ gateways, one balance, privacy architecture.
- [Assistants](https://aizzy.ai/assistants): No-code assistant wizard and 166-skill catalog with SKILL.md import.
- [Roadmap](https://aizzy.ai/roadmap): Planned and shipped features.

## Optional

- [Community](https://aizzy.ai/community): Community links; X: @AizzyAi.